NMR Structure of the Transmembrane Domain of the n-Acetylcholine Receptor β2 Subunit

SUMMARY

Nicotinic acetylcholine receptors (nAChRs) are involved in fast synaptic transmission in the central and peripheral nervous system. Among the many different types of subunits in nAChRs, the β2 subunit often combines with the α4 subunit to form α4β2 pentameric channels, the most abundant subtype of nAChRs in the brain. Besides computational predictions, there is limited experimental data available on the structure of the β2 subunit. Using high-resolution NMR spectroscopy, we solved the structure of the entire transmembrane domain (TM1234) of the β2 subunit. We found that TM1234 formed a four-helix bundle in the absence of the extracellular and intracellular domains. The structure exhibited many similarities to those previously determined for the Torpedo nAChR and the bacterial ion channel GLIC. We also assessed the influence of the fourth transmembrane helix (TM4) on the rest of the domain. Although secondary structures and tertiary arrangements were similar, the addition of TM4 caused dramatic changes in TM3 dynamics and subtle changes in TM1 and TM2. Taken together, this study suggests that the structures of the transmembrane domains of these proteins are largely shaped by determinants inherent in their sequence, but their dynamics may be sensitive to modulation by tertiary and quaternary contacts.

1. INTRODUCTION

The nicotinic acetylcholine receptors (nAChRs) are representative of the Cys-loop superfamily of ligand-gated ion channels responsible for fast synaptic transmission in the central and peripheral nervous system. Muscle-type nAChRs are found at the neuromuscular junction and consist of either (α1)₂β1εδ or (α1)₂β1γδ subunits. Neuronal nAChRs are much more heterogeneous, consisting of a variety of specific co-assemblies of α2-α10 and β2-β4 subunits. α7, α8, α9 and α10 subunits can form homo- or hetero-pentameric receptors with each other; all other α subunits form heteromeric pentamers with β subunits. The β2 subunit is present in the majority of neuronal nAChRs found in the mammalian central nervous system [1, 2] and has been implicated in nicotine addiction [3] and neuroprotection [4]. It assembles with α4 to form α4β2, one of the most abundant nAChR subtypes in the brain. The α4β2 subtype is not only a plausible molecular target for general anesthetics [5-7], but also for other therapeutics in the central nervous system [8, 9].

The overall architecture among nAChR subunits is expected to be similar, given their high degree of sequence homology. Each subunit has an extracellular (EC), transmembrane (TM) and (IC) intracellular domain. The large extracellular domain contains the ligand-binding sites, the transmembrane domain contains four helices (TM1-4) with TM2 facing the pore of the channel, and the intracellular domain is involved with ion channel modulation and cellular localization [10, 11]. All three domains are implicated in ion conductance [9, 12, 13]. Structural models for the muscle-type nAChRs have become available with the cryo-EM structure of Torpedo nAChR (PDB: 2BG9) at 4 Å [14] and the crystal structure of the mouse α1 extracellular domain at 1.94 Å (PDB: 2QC1)[15]. Crystal structures of the homologous bacterial channels GLIC (PDB: 3EHZ; 3EAM) [16, 17] and ELIC (PDB: 2VL0) [18] have also been determined recently. These structures are invaluable to our understanding of the overall architecture of neuronal nAChRs. While nAChR subunits are homologous in sequence and structure, the pharmacological properties of nAChRs comprised by different subunits can vary profoundly. The differences exhibited by various neuronal nAChR subtypes may be reflected in their structural details. Until now, there have been no high-resolution structures available for any neuronal nAChR.

Although cooperative motions of the EC, TM and IC domains of nAChR may be required for normal functioning of the receptor, the structure of one domain may not depend on the presence of the other domain(s). Chimeras replacing the IC domain of mammalian Cys-loop channels with the seven residues of the TM3-TM4 linker in GLIC form robust channels [19], suggesting that the IC domain has minimal influence on TM domain structure. The EC domain may have more influence on the TM domain. Specific contacts between the EC domain and the TM domain at the beginning of TM1, at the loop connecting TM2 and TM3, and at the end of TM4 are thought to be involved in modulation of channel function [20, 21]. However, the major determinants for the structure of a TM domain are likely to be the protein sequence within the domain and the lipid environment surrounding the protein [22, 23]. The TM domains of nAChRs are thus likely to fold largely independent of the EC and IC domains. In fact, the isolated second transmembrane helices (TM2) of the α4 and β2 subunits of human nAChR [24] and the δ subunit of Torpedo nAChR [25] were shown to form functional channels in phospholipid bicelles or vesicles. But the interaction of TM2 with the rest of the transmembrane domain is not trivial. Both mutations in the lipid-facing TM4 helix [26-33] and changes in lipid environment [20, 34] have been shown to affect ion channel assembly and function. The role of TM4 in stabilizing the transmembrane domain is supported by the observation that pentameric assembly of homologous receptors truncated after TM3 were rescued in oocytes by coexpressing the complementary TM4 helix, even though TM4 does not directly form a subunit interface [33]. An understanding of how these factors affect ion channel function requires an understanding of how the transmembrane helices fold together to form a fully functional channel. The feasibility of a reductionist approach to understanding nAChR structure has been established by the crystal structure of the isolated EC domain of the mouse α1 nAChR subunit [15], and by comparing the crystal structures of other membrane proteins to NMR analyses of isolated domains [35, 36].

Here we report an NMR analysis of the entire TM domain of the human β2 nAChR, TM1234. In conjunction with the TM domain of α4, the β2 domain forms the pore of α4β2 nAChR. In the absence of the EC and IC domains, TM1234 folds into a four-helix bundle that more resembles the structure of the TM domain in GLIC [16, 17] than that in the Torpedo β1 subunit [14]. Our study also demonstrated that the presence or absence of TM4 does not greatly affect the secondary structure and tertiary arrangement of the other three TM helices, but without TM4, residues in the other three helices, especially those in TM3, exhibited increased motional flexibility.

2. MATERIALS AND METHODS

2.1 Sample preparation

The regions encompassing transmembrane helices 1-3 (TM123) and 1-4 (TM1234) of the β2 subunit of the human acetylcholine receptor (GenBank: AAB40115.1) were expressed as his-tag fusions in E. coli using the expression vector pTBSG1[37] (a gift from the lab of Professor Timothy Cross at Florida State University). TM123 comprised residues R206-R299 of the mature human nAChR sequence with the native extracellular domain at the N-terminus replaced by 24 amino acids composed of a histidine tag and a TEV protease recognition site. TM1234 comprised residues R206-Y461 of the mature human nAChR sequence, with residues M306-K427 of the intracellular domain replace by a short synthetic linker. Both constructs were optimally expressed in Rosetta 2(DE3) pLysS (Novagen) by induction at 15°C for 5 days using the Marley protocol [38]. Detergent-soluble his-tagged TM123 and TM1234 were routinely purified at about 20 mg per liter of M9 media by solubilization of isolated membranes in 3% empigen and purified on a Ni-NTA resin using standard procedures. Samples were prepared for NMR by removal of the detergent by dialysis and prepared in 50% hexafluoroisopropanol (HFIP)/water with a protein concentration of 0.2-0.25 mM. The anisotropic environment provided by microclusters of HFIP in water proved to be an optimal membrane mimetic for NMR analysis of both TM123 and TM1234. Fluorinated alcohols such as HFIP form amorphous microclusters in water that align with the hydrophobic surfaces of proteins to minimize hydrophobic-aqueous interfaces [39, 40] while allowing interaction between transmembrane helices [41], much as detergents form micelles that accomplish the same effect. It is common practice in preparing membrane proteins for NMR to screen various detergents to establish which detergents prevent aggregation while supporting native structure as reflected by well-dispersed and well-resolved NMR HSQC spectra. This is thought to reflect a good fit between a specific protein and the biophysical properties of the detergent [42, 43]. TM1234 exhibited poor spectral quality and tended to aggregate in all detergents tested, most likely due to the exposed hydrophobic interface normally occupied by the extracellular domain. Hence we chose to use a mixture of HFIP and water as the membrane mimetic instead of a detergent. The degree to which either membrane mimetic sufficiently reflects the complex environment of a lipid bilayer is an open question, but this work confirms that the predicted four-helix bundle of TM1234 can form in HFIP-water mixtures.

2.2 NMR spectroscopy

Bruker Avance 600, 700, 800, and 900 MHz spectrometers were used for all NMR experiments. Each spectrometer was equipped with a triple-resonance inverse-detection cryoprobe (Bruker Instruments, Billerica, MA). Acquisition temperatures between 20°C and 40°C were screened for optimal spectral quality and stability. The highest resolution and signal to noise ratio were obtained at 40 °C. All experiments reported here were acquired at 40°C. The spectral windows in the ¹H and ¹⁵N dimensions were typically set at 12 and 24 ppm, respectively. Relaxation delays of 1.5s (900MHz) and 1s were used. A series of experiments were conducted for chemical shift assignment, including HNCO (1024 × 32 × 80, 800MHz), HNCA and HN(CO)CA with a ¹³C spectral width of 28 ppm (1024 × 32 × 80, 700MHz), and CBCA(CO)NH with a ¹³C spectral window of 60 ppm (1024 × 32 × 80, 700MHz). 3D ¹⁵N-edited NOESY and ¹³C-edited NOESY spectra were acquired with mixing times of 150, 180, or 200 ms. In addition, 2D ¹H homonuclear NOESY with 1024 × 400 data points and a 150 ms mixing time was performed at 700 MHz to assist in chemical shift assignments, particularly those of the ring protons of aromatic residues such as tyrosine, phenylalanine, and tryptophan. HSQC spectra were collected as 1024 × 128 data points. The observed ¹H chemical shifts were referenced to the DSS resonance at 0 ppm and the ¹⁵N and ¹³C chemical shifts were indirectly referenced [44].

2.3 Data Process and Analysis

NMR data were processed using NMRPipe 4.1 and NMRDraw 1.8 [45], and analyzed using Sparky 3.10 [46]. Chemical shift assignment for TM123 was performed manually using HNCO, HNCA, HNCOCA, and CBCACONH, as well as NOESY spectra. The chemical shifts for TM1234 were determined based on the results from TM123 and an additional set of NMR experiments. A total of 100 structures were calculated by CYANA 2.1 [47] based on NOE restraints and Talos dihedral angle restraints from the chemical shift index (CSI) [48]. A final bundle of 20 structures with the lowest target function were analyzed using VMD [49] and Molmol [50].

Contact areas between residues were analyzed using the CMA component of the SPACE suite [51]. Contact surface area was defined as the area between two atoms into which a solvent molecule cannot fit. Average atomic packing was analyzed using Voronoia [52]. Average atomic packing density was calculated using a Voronoi Cell algorithm and defined as the assigned atomic volume inside the atom’s van der Waals radius divided by the sum of this volume and the remaining solvent excluded volume. Surface pockets and interior voids were calculated using CASTp [53] with a probe radius of 1.4 Å. Voids were defined as enclosed empty space inside proteins inaccessible to water molecules from outside. Pockets were defined as crevices with constricted openings but accessible to water molecules from outside. For this study, axial volume was defined as the sum of the volumes of the empty space along the central axis of the transmembrane 4-helix bundle.

3 RESULTS AND DISCUSSION

3.1 TM1234 and TM123 structures

Figure 1 shows typical [¹⁵N,¹H]-HSQC NMR spectra of TM123 and TM1234, where 98% of the residues were well-resolved and assigned. The complete assignments are shown in Supplementary Materials, Figure 1S and 2S. The secondary structures of TM123 and TM1234 were determined by analyzing C_α chemical shifts, indicating a helical content of 69% and 64%, respectively. Note that many resonance peaks of TM123 overlapped well between the two spectra, indicating that TM123 folds into similar structures in the absence and presence of TM4.

Fig. 1.

Overlay of [¹⁵N,¹H]-HSQC NMR spectra of TM123 (red) and TM1234 (black) of the human nAChR β2 subunit reveals overall similarity. Some peaks common to both TM123 and TM1234 showed changes in chemical shift and peak intensity in the presence and absence of TM4. Chemical shift assignments can be found in the supplementary materials.

The structure of TM1234 was calculated based on chemical shifts, 1427 short and medium-range NOEs, and 27 long-range interhelical NOEs (Figure 2). Several long-range representative NOE’s demonstrating interhelical connectivities are shown in Figure 3S. Table 1 summarizes the statistics for the structure calculation. Figure 3A depicts a bundle of the 20 lowest-energy structures of TM1234 (PDB ID code 2KSR, BMRB: RCSB101528). The averaged pair wise root mean square deviation (RMSD) calculated for backbone-only and for all heavy atoms in the helical regions are 0.94 and 1.24 Å, respectively. Similar data for TM123 are shown in Table 1S and Figure 4S. TM1234 folded into a four-helix bundle. Each helix is composed of 22 to 29 residues. The average atomic packing density was 0.788, indicating a tightly packed structure [52]. The tertiary structure was defined primarily by hydrophobic contacts between helices, with an interhelical contact surface area [51] of 1447 Ų involving 65 pairs of residues. Bulky amino acids, such as PHE, LEU and ILE, dominated the hydrophobic contacts with no apparent packing motif. The greatest contact areas were between TM3 and TM4 and between TM1 and TM2, each contributing about 30% of the total contact area between helices. Other adjacent TM helices contributed about 16% for each pair. There was no contact between opposed TM2 and TM4 helices, but the opposed helices of TM1 and TM3 contributed 7% of the total helical contacts. In the absence of TM4, TM123 folded into a structure similar to that of TM1234. The structural overlay of TM123 and TM1234 is shown in Figure 3B. The backbone RMSD of TM123 in the absence and presence of TM4 is 2.4 Å.

Fig. 2.

NOE connectivity and C_α chemical shift index of TM1234. Sequential, midrange and long-range NOE connectivities are linked by line segments with widths proportional to the observed NOE intensities. The helical regions of the calculated protein structure are indicated below the sequence.

Table 1.

Statistics for the family of 20 calculated structures of the TM1234 domain of the human nAChR β₂ subunit in 50% HFIP/50% H₂O.

NMR structure	Statistics
Number of distance restraints	1454
Intraresidue (∣i – j∣ = 0)	524
Short range (∣i – j∣ = 1)	514
Medium range (1 < ∣i – j∣ ≤ 4)	389
Long range, interhelical (∣i – j∣ ≤ 5)	27
Number of dihedral angle restraints  (Residues 29-37, 39-51, 57-79, 90-115, 131-158)	198
Number of upper limit restraints violations > 0.5 Å	0
Number of dihedral angle restraints violations > 5°	0
Backbone RMSD (Residues 29-50, 57-78, 88-116, 134-157)	0.94 ± 0.27 Å
Heavy atom RMSD (Residues 29-50, 57-78, 88-116, 134-157)	1.24 ± 0.27 Å
Ramachandran plot
Residues in most favored regions	87.1%
Residues in additionally allowed regions	13.2%
Residues in generously allowed regions	0.3%
Residues in disallowed regions	0.3%

Fig. 3.

(A) Bundle of 20 lowest-energy structures of TM1234. Structures were calculated from NOE and Talos dihedral angle restraints using CYANA 2.1. Backbone RMSD for the helical regions is 0.94 ± 0.27 Å. Further statistics for the structure calculation are summarized in Table 1. (B) Comparison of the structures of TM123 (yellow) and TM1234 (color scale from red (TM1) to blue (TM4)) demonstrating the effect of TM4 on TM123 structure. The backbone atoms of the helices aligned with an RMSD of 2.4°A.

The TM1 helix comprised residues L29 to F50, with a kink at residue L35 evident from the C_α chemical shift index in Figure 2. The disruption of helical structure at this site was also observed in our previous NMR experiments on the isolated TM1 helix in DPC detergent, where the helical structure extended from L35 to F50 [54]. Thus, the helical kink at L35 seems to be an intrinsic property of the protein and independent of the membrane mimetic. The lack of helical structure before L35 in the isolated TM1 helix may reflect a difference in the mimetic environment, but may also be influenced by the absence of interaction with the other helices of the 4-helix bundle. The kink is likely due to the presence of an evolutionarily conserved proline three residues downstream [55, 56]. While not visible in the relatively low-resolution structure of Torpedo nAChR [14], homologous residues in the crystal structures of both ELIC and GLIC show similar kinks [16-18]. Accessibility studies of the pre-TM1 region in the homologous GABA_A receptor in the presence and absence of ligand indicated movement of this region on ligand binding [57]. This seemingly conserved kink may act to allow a change in helical angle for optimal packing.

Pore-lining residues in TM2 (T61, S65, L68, V72 and L76) suggested previously by homology to the Torpedo β1 subunit [58, 59] are facing the same side of the TM2 helix in the TM1234 structure. The secondary structure of TM2 in TM1234 is similar to that obtained from our previous NMR experiments on the isolated TM2 peptide (residues E58 to S85) in DPC detergent, except that in the TM1234 structure the helix begins at G57, before the start of the TM2 peptide [60].

The TM2-3 loop contains residues 79KIVPPTSLD87. Residues KIVP extend along the axis of the TM2 helix; residues PT form a bend at the top of the loop, followed by residues SLD and the beginning of the TM3 helix. Cis-trans isomerization of proline in the TM2-3 loop has been proposed to be involved in the coupling of ligand binding to channel gating [61]. All prolines in TM1234 are in the trans conformation.

3.2 TM4 modulates the motion of other helices

Despite the fact that the structure of TM123 remained more or less the same in the absence and presence of TM4, as revealed in Figure 1, the influence of TM4 on the other helices does exist, especially on TM3. As shown in Figure 4, several residues at the TM3 c-terminus exhibited profound changes in their chemical shifts, largely due to the change from a free terminus to a region linked with TM4. A more interesting observation was that the NMR resonance intensity of TM3 residues became much weaker when TM4 was present. Changes in peak intensity can be related to changes in backbone dynamics (for example, see Fig. 8s in [62]). In the absence of TM4, residues in TM3 showed stronger resonance intensities than those residues in TM1 and TM2, suggesting greater mobility for residues in TM3. In the presence of TM4, the overall intensity of TM3 resonances was reduced, indicating that the mobility of TM3 was restrained by TM4. Even though TM4 is also in direct contact with TM1, the presence of TM4 had much less effect on TM1 than TM3, presumably because TM1 was intrinsically less mobile than TM3. Thus it appears that the presence of TM4 caused only subtle changes to the structure, but larger changes to the dynamics of TM1234.

Fig. 4.

Effect of TM4 on TM123. Changes in chemical shift and intensity for each residue in the absence and presence of TM4 are indicated. Changes in chemical shift were (A) plotted as a function of residue number and (B) color-coded onto the TM1234 structure with a color scale from blue (no effect) to white (chemical shift changed ≥ 0.3 ppm ). Changes in peak intensity were (C) plotted as a function of residue number and (D) color-coded onto the TM1234 structure with a color scale from blue (no effect) to white (intensity reduced to zero). For clarity, the TM4 helix is not shown. The change in chemical shift (Δδ(H+N) was calculated as $\Delta \delta (H+N)={\left(\right(\Delta {\delta }_H^2+(\Delta {\delta }_N^2∕25))∕2)}^{1∕2}$.

TM2 is not in direct contact with TM4 (see Figure 6 below), but it showed mild changes in chemical shift and intensity for multiple specific residues throughout the entire helix upon the addition of TM4 (Figure 4). For example, V72 in TM2 (V253 in the full-length structure), believed to act as a hydrophobic girdle in the pore of the pentameric assembly [58, 59], showed a 55% reduction in peak intensity after addition of TM4. Thus, TM4 affected motional properties of residues not only in direct contact with TM4 but also remote from the TM4 interface. The ability of TM4 to allosterically modulate the dynamics of the other helices may help explain previous observations that mutations in TM4 affect channel functions [26-32]. Our data are also in agreement with molecular dynamic studies predicting the influence of the TM4 helix on the Torpedo nAChR transmembrane domain [63-66]. Small influences on domain structure and dynamics may prove profoundly important for protein function. There is strong experimental evidence that changes in protein motions may activate allosteric proteins that are otherwise structurally inactive (see, for example, [67]). Molecular dynamics studies of nAChR predict changes in dynamics can account for changes in channel function [58, 65, 68-71].

Fig. 6.

Contact map analysis. Contact between residues was calculated using CMA for (A) TM1234 and (B) the 2BG9 β1 subunit. TM1234 shows greater interaction between helices TM1 and TM4.

3.3 Structural comparisons of TM1234 with homologous proteins

Among the known homologous structures, the transmembrane domain of human β2 nAChR shares 75% homology (62% identity) with the β1 subunit of Torpedo nAChR, 41% homology (22% identity) with GLIC, and 37% homology (20% identity) with ELIC. It is anticipated that TM1234 of the β2 nAChR shares structural similarities with these proteins. The superimposed structures of TM1234 upon the transmembrane domain of the β1 subunit of Torpedo nAChR and the GLIC transmembrane domain are provided in Figure 5. Comparison of our TM1234 structure with the structures of corresponding regions in the β1 subunit of Torpedo nAChR, GLIC and ELIC showed a helical backbone RMSD of 5 Å, 4.4 Å and 5.9 Å, respectively. For a more detailed comparison, we also analyzed the average atomic packing densities, surface areas of residue-to-residue contacts, and packing defects on each of these structures. The results are summarized in Table 2. By all three measures, TM1234 is intermediate between the structures of GLIC and the Torpedo nAChR β1 subunit.

Fig. 5.

Comparison of TM1234 with homologous structures. The backbone atoms of the transmembrane helices for each homologous protein (white) were aligned with TM1234 (blue). (A) 2BG9 β1 subunit (RMSD 5 Å). (B) GLIC (RMSD 4.4 Å).

Table 2.

Packing efficiency in TM1234 and the TM domains of homologous structures. For a description of these measures of packing, see Methods, section 3.2.

Measure of packing	TM1234	GLIC	β1 Torpedo nAChR
Average atomic packing densities	0.788	0.750	0.802
Surface area of residue-to-residue contacts (Å2)	1447	2328	1428
Packing defect as axial volume (Å3)	2056	1009	3416

Detailed inter-helical contacts in TM1234 and in the Torpedo nAChR β1 subunit are elucidated in Figure 6. The most noticeable difference between TM1234 and the Torpedo nAChR β1 subunit is that the c-terminal end of TM4 had almost no contact with TM1 in the Torpedo nAChR β1 subunit, but had extensive contacts in TM1234. Lack of interactions with the missing extracellular domain in TM1234 may have promoted interactions between the end of TM4 and the beginning of TM1 that stabilize the four-helix bundle.

Despite an overall structural similarity, these structures deviate significantly from each other at the extracellular interface, especially at the TM2-3 loop. For TM1234, the structural deviation in this region appears to be an intrinsic property of the sequence. The same region of GLIC also differs substantially from that of the Torpedo nAChR β1 subunit, in which the TM2 and TM4 helices extend further into the extracellular interface than they do in GLIC. The secondary structure of TM1234 in this region more closely resembles GLIC, with TM2 ending a few residues before the conserved proline. Relative to the Torpedo nAChR β1 subunit, the TM2 helix of TM1234 begins two residues earlier than the homologous region of the Torpedo nAChR β1 subunit, but ends seven residues earlier. Even so, the residues predicted to be lining the pore at the end of TM2, while not technically in the TM2 helix, are positioned such that they are still accessible to the pore. An earlier helical termination at the c-terminus of TM2 was also observed in our previous NMR structure of TM2 in DPC micelles [60], suggesting that a shorter TM2 helix may result from intrinsic sequence determinants regardless of the membrane mimetic.

4 Conclusions

It is intriguing to see that TM1234 alone under our experimental conditions could fold into a structure that so closely resembles the structures of the TM domains of other intact homologous proteins. It is also interesting to see that the structure of TM1234 in water/HFIP resembles well the structures of homologous proteins determined in detergents and lipids. Taking all the results from this study together, at least three notions can be derived that may have general implications. First, secondary structure (not only helices but also non-repetitive features such as kinks) can be preserved in a reductionist approach to structural analysis. Secondly, as long as a membrane mimetic environment is provided, a robust transmembrane domain such as TM1234 can fold independently. The membrane mimetic or the associating extracellular and intracellular domains may influence some details of transmembrane domain structure, but the overall structure is largely defined by intrinsic sequence determinants within the domain. These two realizations assure the feasibility of a reductionist approach for dissecting structural information of large protein assemblies. These results are consistent with other studies using a reductionist approach to analyze large proteins, such as the Sec translocase [72], and membrane proteins such as the lactose permease and bacteriorhodopsin [35, 36]. Our third notion is that the dynamic properties of a multi-domain protein are dependent on the interplay between domains. Even within the transmembrane domain, the addition of TM4 had little effect on TM123 structure, but its effect on backbone dynamics extended beyond residues that had direct contact with TM4. Collectively, the reductionist approach may be adequate for structural analysis, but it is probably inadequate to reveal important dynamical information of proteins.